The proliferation of industrial robotic systems in manufacturing and assembly is well known and becoming very significant. Many robotic systems are assembled from off-the-shelf standard electrical/mechanical components. Many of these components are bulky and heavy and because the resulting robotic system is usually designed around these components, this results in unnecessary design constraints. For example, many industrial robotic joints are driven by electric motors mounted away from the joint and connected to the joint by a long rod. These types of robots suffer from a limited range of motion of the robot arm and such configurations must typically be used in dedicated applications. In many environments requiring robotic assistance, there are many different tasks to which a given robot could be assigned. However, such an advantage is not realized due to the aforementioned design constraints which precludes convenient conversion of the robot from one application to another. It would therefore be very advantageous to have a robot designed in such a way that it is free of undue design limitations and readily adapted to a variety of tasks. In other words, there is a need for a robotic system which is modular, reconfigurable and expandable.
The need for new robotic systems is also motivated by the lack of versatile test beds and dedicated systems useful for grasping and manipulation. Existing systems such as the Utah-MIT hand, the Stanford/JPL hand and others are mostly anthropomorphic and well-suited for specific applications only, have a small workspace and limited dexterity, or are mechanically dependent i.e. they require a supporting manipulator. There have also been robots constructed using a pair of similar manipulators for grasping an object with fixed contact, producing results of limited scope. Because there has been very little work done in this area of robotics it is not clear what types of mechanisms are most suitable. In order to provide a more versatile robotic system which overcomes the aforementioned limitations it is necessary to address issues such as (i) the suitability of a given kinematic structure for a given manipulation task; (ii) intelligent use of multiple manipulators for complex tasks such as parts-mating: (iii) compatibility of individual manipulators with different kinematic structures and dynamic properties: (iv) testing of cooperation schemes, and (v) optimization of load distribution to name a few.
The performance of the mechanical robotic system with many degrees of freedom is critically dependent on the real-time computer control system. Such a system must sample all the outputs of all the various types of sensors, compute the control commands dictated by the user-defined control algorithm and send the commands out to the actuators all within a pre-defined sampling interval. A drawback of existing computer controllers for multifingered hands and multiple manipulator systems such as those disclosed in Bejczy, A. and Szakaly, Z., "Universal Computer Control System (UCCS) for Space Telerobots", Proc. of IEEE Int. Conf. on Robotics & Automation, pp. 318-324, Raleigh, 1987; and Narasimhan, S., Siegel, D M. and Hollerbach, J. M., "Condor: A Revised Architecture for Controlling the Utah/MIT Hand", Proc. of IEEE Int. Conf. on Robotics & Automation, pp. 446-449, 1988; is that they have been designed intuitively starting from the available microprocessor technology without quantifying requirements in terms of computational complexity, sampling rate and I/O speed.
Therefore, for large-scale robotic systems instrumented with discrete and distributed sensors and actuators it would be advantageous to provide a real-time computer architecture capable of high computational throughput that can accommodate sophisticated control laws including complex feedback compensation and adaptive strategies, sub-millisecond sampling periods and high speed digital and analog I/O.